Mechanistic modeling of bituminous mortars to predict performance of asphalt mixtures containing RAP

MECHANISTIC MODELING OF BITUMINOUS MORTARS TO PREDICT

PERFORMANCE OF ASPHALT MIXTURES CONTAINING RAP

Dissertation

submitted to and approved by the

Department of Architecture, Civil Engineering and Environmental Sciences

University of Braunschweig – Institute of Technology

and the

Department of Civil and Environmental Engineering

University of Florence

in candidacy for the degree of a

Doktor-Ingenieurin (Dr.-Ing.) /

Dottore di Ricerca in Civil and Environmental Engineering

by

Chiara Riccardi

born 05/01/1987

from Viareggio (LU), Italy

Submitted on

21.02.2017

Oral examination on

08.05.2017

Professorial advisors

Prof. Massimo Losa

Prof. Michael P. Wistuba

i

Abstract

In last decades the use of Reclaimed Asphalt Pavement (RAP) materials in asphalt mixtures has seen a significant expansion for economical and environmental reasons. Nevertheless, there are still two important issues which have not been effectively solved: the first regards the characterization of the aged binder contained in RAP; the second concerns the ability to predict the performance of asphalt mixture composed with high RAP content.

Regarding the first problem, the current methods are based on extraction and recovery of the RAP binder using solvents; however, these methods are not fully accurate since they can alter the rheological properties of the binder. For this reason, in the present work, a new procedure to back-calculate the rheological properties of the aged binder contained in RAP materials and of the blends composed with fresh and RAP binder, was developed. This is based on DSR tests performed on mortars, composed by mixing the fine fraction of the RAP aggregate with virgin binder. Using the Nielsen model, specifically adapted to asphalt mortars’ case and the Voigt model, the rheological properties of RAP binder can be back-calculated from mortars tests. The present procedure has two advantages: the aged binder contained in RAP is tested as it is after the milling process, avoiding any further treatments, while the testing campaign is centered on the mortar phase, which is one of the most important phases governing the properties of the mixtures and, therefore, the performance of asphalt pavements.

Regarding the second problem, a multi-scale approach based on rheological model (2 Spring, 2 Parabolic Elements, 1 Dashpot, 2S2P1D model) and on empirical models (Hirsh e Witczak models) was developed to predict the rheological properties of asphalt mixture containing RAP materials, starting from tests on asphalt mortars and taking into account the grading of the aggregates and the volumetric composition of the mixtures.

This methodology allows to make reliable previsions for both the problems addressed, as demonstrated by the results of the validation tests carried out in this doctoral thesis. In addition, the present research provides innovative solutions to address some of the issues which are currently of particular importance for the purposes of extending the use of RAP material in the production of asphalt mixture.

Key words: RAP, mortar, blend, rheology, Nielsen model, Voigt model, 2S2P1D model, Hirsh model, Witczak model.

ii

Abstract

Nell’ultimo ventennio si è assistito a un rapido incremento dell’utilizzo di materiale fresato nelle pavimentazioni stradali dovuto in parte al risparmio economico derivante dall’utilizzo di questo materiale e in parte ai vantaggi ambientali. Tuttavia, esistono ancora due importanti problematiche alle quali non è stata data una risposta efficace: la prima riguarda la caratterizzazione del bitume invecchiato contenuto nel fresato; la seconda riguarda la possibilità di prevedere le prestazioni delle miscele confezionate utilizzando elevate quantità di fresato.

Con riferimento alla prima problematica, i metodi attualmente disponibili sono basati sull’estrazione e sul recupero del bitume contenuto nel fresato per mezzo di solventi. Tali metodi presentano però vari problemi connessi al fatto che l’estrazione e il recupero del bitume può alterarne le proprietà reologiche. Per questo motivo, nel presente lavoro, si è sviluppata una nuova procedura che permette di eseguire un retrocalcolo delle proprietà reologiche del bitume invecchiato e dei blends, composti da bitume invecchiato e bitume vergine, partendo da test su malte composte miscelando la parte fine degli aggregati del fresato e il bitume vergine; utilizzando il modello di Nielsen, appositamente adattato al caso di materiali bituminosi, e il modello di Voigt si riesce a determinare le suddette proprietà reologiche utilizzando i risultati dei test eseguiti sulle malte. Ciò ha un duplice beneficio: da una parte è possibile caratterizzare il bitume invecchiato contenuto nel fresato nelle stesse condizioni di lavoro in cui si trova all’interno della pavimentazione, evitando che subisca ulteriori trattamenti, dall’altra si possono eseguire le prove di caratterizzazione su una delle fasi che maggiormente governano le proprietà delle miscele in conglomerato bituminoso e che quindi regolano le performance delle pavimentazioni.

Con riferimento alla seconda problematica, è stata sviluppata una procedura di modellazione multiscala, basata sia su modelli reologici (2 Spring, 2 Parabolic Elements, 1 Dashpot, 2S2P1D model) sia su modelli empirici (Hirsh e Witczak models), che permette di prevedere il comportamento reologico delle miscele in conglomerato bituminoso, contenenti materiale fresato, partendo direttamente dai risultati di prove eseguite sulle malte e tenendo conto della composizione granulometrica degli aggregati e di quella volumetrica delle miscele.

La metodologia messa a punto consente di ottenere previsioni affidabili per entrambe le problematiche affrontate, come dimostrato dai risultati delle prove di validazione eseguite nell’ambito della tesi, e di fornire una soluzione innovativa ad alcune questioni che risultano attualmente di particolare rilievo ai fini della estesa utilizzazione del fresato per il confezionamento dei conglomerati bituminosi.

Parole chiave: fresato, mortar, blend, reologia, modello di Nielsen, modello di Voigt, modello 2S2P1D, modello di Hirsh, modello di Witczak.

iii

Abstract

In den vergangenen Jahrzehnten hat der Einsatz von Ausbauasphalt beziehungsweise Reclaimed Aspalt Pavement (RAP) in Asphaltmischgut aus ökonomischen und ökologischen Gründen eine deutliche Ausweitung erfahren. Jedoch gibt es bezüglich zwei wichtiger Aspekte, noch keine effektiven Lösungsansätze: Der erste Aspekt ist die Charakterisierung des im RAP enthaltenen gealterten Bitumens; Der zweite Aspekt betrifft die Prognose der Gebrauchseigenschaften von von Asphalt mit einem hohen Zugabeanteil von RAP.

Hinsichtlich des ersten Problems beruhen die derzeitigen Verfahren auf der Extraktion und Wiedergewinnung des RAP-Bindemittels unter Verwendung von Lösemitteln; Diese Verfahren sind jedoch nicht hinreichend genau, da sie die rheologischen Eigenschaften des Bindemittels verändern können. Aus diesem Grund wurde in der vorliegenden Arbeit ein neues Verfahren zur Rückrechnung der rheologischen Eigenschaften des in RAP-Materialien enthaltenen gealterten Bindemittels und der Verschneidung des frischen und des RAP-Bindemittels entwickelt. Dieses Verfahren basiert auf DSR-Analytik, die an Asphaltmörteln durchgeführt wurde. Die Asphaltmörtel wurden aus deraus der feinen Anteile des RAP und frischem Bindemittel zusammengesetzt. Auf Basis des Voigt-Modells in Kombination mit dem Nielsen-Modell, das speziell bezüglich der Asphaltmörteleigenschaften angepasst wurde, können die rheologischen Eigenschaften des RAP-Bindemittels aus den Ergebnissen der DSR-Analytik zurückberechnet werden. Das vorliegende Verfahren bietet zwei Vorteile: Das im RAP enthaltene gealterte Bindemittel kann direkt nach dem Fräsvorgang getestet werden ohne weitere Konditionierung. Weiterhin konzentrieren sich die Untersuchungen an der Asphaltmörtelphase, welche maßgebenden Einfluss auf die Asphaltmischguteigenschaften und den resultierenden Gebrauchseigenschaften hat.

In Bezug auf das zweite Problem wurde ein Mehrskalen Modell angewandt, das auf einem rheologischen Modell (2 Spring (Federn), 2 Parabolische Elemente, 1 Dämpfer, 2S2P1D Modell) und auf empirischen Modellen (Hirsh e Witczak Modelle) basiert, um die rheologischen Eigenschaften des Asphalts zu prognostizieren. Die Eingangsparameter des Mehrskalen Modells werden mittels DSR-Analytik am Asphaltmörtel unter Berücksichtigung der Korngrößenverteilung und der volumetrischen Zusammensetzung bestimmt.

Die Validierungsergebnisse dieser Arbeit zeigen, dass dieentwickelte Methodik eine zuverlässige Prognose der angesprochenen Aspekte ermöglicht. Darüber hinaus bieten die referierten Ergebnisse innovative Lösungsansätze , um die Verwendung von RAP in der Asphaltmischgutproduktion zu maximieren.

Schlüsselwörter: RAP, Asphaltmörtel, Rheologie, Nielsen Modell, Voigt Modell, 2S2P1D Modell, Hirsh Modell, Witczak Modell.

iv

Acknowledgements

I would like to express my gratitude towards a number of individuals who have helped make this research possible. First, my two advisors, Prof. Massimo Losa and Prof. Michael Wistuba, which gave me the possibility to do this incredible experience that really changes my life. Thanks for guiding and supporting me during these three years, also in the tough moments.

I would like to thank Prof. Pietro Leandri and Ing. Patrizia Rocchio for their support and advices during the years.

I would like to thank Dr. Augusto Cannone Falchetto for his precious support and help.

I would like to thank all the ISBS team for helping me with the huge amounts of laboratory tests, and also for their support during my stay in Braunschweig.

I would like to thank all my friends (the old and the new one) with whom I share many of the joys and challenges of these three years.

I would like to thank my family for their love, support, and constant encouragement over the years. In particular, I would like to thank my brother for always staying close to me even if he was physically far away.

Lastly, and most importantly, I would like to thank my future husband, Gaspare, for his love, support, and patience throughout this process, without which I most certainly would not have succeeded. Finally, I would like to thank and dedicate this thesis to my mother, Giovanna. Although it has been years since you have passed, I still take your teaching and your dreams with me, every day.

Index

v

Index

Abstract (in English) ... i

Abstract (in Italian) ... ii

Abstract (in German) ... iii

Acknowledgements ... iv

Index ... v

Chapter 1 ... 1

1.

Introduction ... 1

1.1 Context ... 1 1.2 Objective ... 1

1.3 Overview of the methodology ... 2

1.4 Innovative Aspects ... 3

1.5 Practical application of the procedure ... 4

Chapter 2 ... 5

2.

Literature review... 5

2.1 Existing models ... 5

E.1 2S2P1D Model ... 5

E.2 Empirical Models to determine the complex modulus of mixtures from the complex modulus of binders ... 8

E.2.1 Hirsch model ... 8

E.2.2 Witczak 1-40D model ... 9

E.3 Nielsen model... 9

E.4 Voigt Model ... 12

E.5 Arrhenius Model ... 13

2.2 Master Curves ... 13

2.2.1 Christensen Anderson Maresteanu (CAM) model... 13

2.2.2 Sigmoidal model ... 14

Chapter 3 ... 15

3.

Estimation of the rheological properties of RAP binders avoiding the extraction and

recovery method ... 15

3.1 Development of a new procedure to estimate the RAP binder properties from results on mortar tests ... 15

3.1.1 Development of the Enhanced Nielsen model (N.1) ... 16

3.1.2 Description of materials and tests ... 18

3.1.2.1 Artificial RAP source ... 20

3.1.2.2 Italian RAP source ... 24

3.1.2.3 English RAP source ... 31

3.1.2.4 German RAP source... 36

Index

vi

3.1.3.1 Calibration of the Enhanced Nielsen model parameters on BSRAP mortar ... 42

3.1.3.2 Extension of the calibrated Enhanced Nielsen model to SRAP mortars ... 45

3.1.3.3 Application of the Voigt model to estimate the rheological properties of RAP binder ... 47

3.1.3.4 Global validation of the procedure ... 50

3.1.3.5 Application of the procedure to Italian RAP source ... 53

3.2 Implementation of a new procedure to back-calculate the Performance Grade (PG) of binders from Master Curves ... 55

3.2.1 High Temperature PG ... 57

3.2.2 Low Temperature PG ... 59

3.3 Determination of the maximum amount of RAP that can be added in a mixture without compromising its performance ... 63

3.3.1 Blending charts to determine the maximum amount of RAP ... 63

3.3.2 Analytical procedure to determine the maximum amount of RAP ... 65

Chapter 4 ... 69

4.

Implementation of the 2S2P1D Model for multiscale modeling of asphalt mixtures ... 69

4.1 Extension of the 2S2P1D model to binder/mortar (N.2) ... 69

4.1.1. Materials and Tests ... 70

4.1.2 Calibration of the 2S2P1D model ... 72

4.1.3 Relationship between the characteristic times of binder and of mortar ... 74

4.1.3.1 Influence of RAP percentage ... 75

4.1.4 Validation ... 78

4.2 Extension of the 2S2P1D model to mortar/mixture (N.3) ... 82

4.2.1. Materials and Tests ... 83

4.2.1.1 Asphalt mixtures ... 83

4.2.1.2 Asphalt mortars and asphalt blends ... 89

4.2.2Calibration of the 2S2P1D model ... 89

4.2.3Relationship between the characteristic time of mortar and of mixture ... 90

4.2.4 Validation ... 91

4.3 2S2P1D Model linking binder/mixture phases ... 92

4.3.1. Materials and Tests ... 93

4.3.2 2S2P1D model fitting and determination of the α parameter ... 96

4.3.3 Multiple-regression analysis of parameter α ... 98

4.4 Summary ... 99

Chapter 5 ... 103

5.

Empirical models ... 103

5.1 Hirsch model ... 103 5.2 Witczak model ... 106 5.3 Summary ... 109

Chapter 6 ... 113

6.

Conclusion ... 113

References ... 115

ANNEX 1 ... 123

Chapter 1

1

Chapter 1

1. Introduction

1.1 Context

In the last decades, the use of Reclaimed Asphalt Pavement (RAP) materials in asphalt mixtures has seen a significant expansion due to the increase of materials costs, and to a deeper understanding for a more environmentally oriented construction process (Kennedy et al., 1998; Holtz and Eighmy, 2000, McGraw et al., 2010; Wistuba et al., 2012; Radenberg et al., 2012, Cannone Falchetto et al., 2012). The use of RAP can increase the sustainability of asphalt pavements by reducing the use of new asphalt binder and virgin aggregates and by limiting material disposal (Hansen and Copeland, 2014). Despite the massive use of RAP in the production of asphalt pavement, an accurate estimation of the rheological properties of the aged binder contained in RAP materials is still a challenge. In literature phenomenological and theoretical relations are available for evaluating performance of asphalt mixtures starting from binder properties. Yet, relevant problems arise when these approaches are used for RAP that are related mainly to difficulties in determining properties of the aged RAP binder. Current methodologies used to determine properties of the RAP binder are based on the extraction and recovery method, as specified in EN 12694, EN12697 and in AASHTO T164, but many research studies (Stroup-Gardiner and Nelson, 2000; Ma and Zhang, 2008; Ma and Huang, 2008) consistently showed that this method is not accurate for many reasons: 1) it alters the binder properties; 2) the solvent extraction produces binder hardening, as shown by Kondrath, 2009 and Burr et al., 1991. Further concerns are associated to the presence of residual solvent after recovery and to the further aging of binder during the heating process. For example, researches indicated even 0.5% residual solvent could cause a 50% decrease in viscosity (Peterson et al., 2000).

These problems considering, its difficult to predict reliably mixture performance from properties of the blend of fresh and RAP binders, and, therefore, its difficult to determine the maximum percentage of RAP, without elaborate asphalt mixture testing, that can be added into a mixture without compromising its rheological and mechanical performance.

1.2 Objective

In order to overcome the limitations listed in the previous section, this thesis focuses on an analytical forward and inverse modeling with the aim of predicting RAP-modified asphalt materials performance across the different material phases. Asphalt mixture is commonly modeled as a composite material consisting of three constituents: air voids, asphalt binder and aggregate of different size and shape. In

Chapter 1

2

the past and more recently, different research efforts have conventionally identified a number of phases within asphalt mixture: asphalt mastic, mortar, fine aggregate matrix (FAM). In particular, asphalt mastic contains fillers ranging from 1μm to 75μm in a binder matrix; mortar containing fine aggregate particles from 75μm to 150μm in a mastic matrix (Arshadi et al., 2014); and FAM are composed of fine aggregate particles smaller than 1.18 mm or 2.36 mm, depending on the Nominal Maximum Aggregate Size (NMAS) of the mixture (Underwood et al., 2013), in a mortar matrix. The goal of the present thesis is achieved through the possibility of experimentally measuring the properties of binders, mortars and mixtures, while avoiding extraction and recovery of the RAP binder, and by rheological modeling the interrelation between the different material phases (a detailed explanation of the methodology is presented in the following section).

This approach will be validated through a preliminary investigation of different mechanical properties for limited types of material factor-level combinations (such as particles content, binder type, RAP source, aging levels).

The research plan consists in the development of a new procedure, that is based on a multi-phase approach, for predicting the mechanical properties of asphalt mixture, such as Complex modulus E* and phase angle δ, from those of binders and mortars; particularly, these latter are composed of a selected fine fraction of RAP (passing the 0.15 mm sieve) and fresh binder. Based on this procedure, the extraction and recovery method is avoided, and the RAP binder contained in the selected fraction of the RAP can be used directly for testing, avoiding any further type of treatment (solvent extraction, oxidation). Moreover, tests on asphalt mixtures samples that are costly and time consuming, can be avoided or considerably reduced.

The proposed procedure can be used also in the inverse mode for determining properties of binders directly from tests on mortars; by this way, existing models, specifically adapted, can be used directly to predict the mechanical properties of mixtures from binder properties avoiding the extraction and recovery of the RAP binder.

1.3 Overview of the methodology

The methodology used in the present study is summarized in the flow chart of Figure 1; letter E identifies the existing models while letter N indicates the newly proposed linking relationships between the different material phases, which are addressed in this thesis.

Chapter 1

3

E: Existing model; N: New proposed links between the different material phases.

Fig.1 Research methodology

Specifically, as shown in Figure 1, two different approaches can be used to determine the mechanical properties of mixtures: one based on rheological models and one relying on empirical models.

The rheological 2S2P1D (2 Springs, 2 Parabolic elements, 1 Dashpot) model, developed by Di Benedetto et al., 2004, links the mechanical properties of mixtures to those of binders (E.1 in Figure 1). In the present study, the use of this model will be extended to the prediction of the rheological properties of mortar from those of binder and to the estimation of the response of mixture from that of mortar (link N.1, N.3 in Figure 1). At the same time, the variation of the characteristic time as function of RAP binder percentage will be analyzed, since this 2S2P1D model parameter governs the temperature dependency and representing the time needed for the system to relax. Furthermore, a link between the characteristic time of the mixture and its volumetric composition will be sought.

Two additional models, the empirical Witczak and micro-mechanical Hirsch models (E.2 in Figure 1) (Bari et al., 2006; Christensen et al., 2003), which allow predicting the complex modulus of mixtures from those of binders, will be used. In the present work, the Witczak model (Bari et al., 2006) will be calibrated in order to predict complex modulus of mixtures from those of the corresponding mortars replacing the traditional binder data input (N.3 in Figure 1).

Moreover, a new procedure for estimating binder properties by performing tests on mortar will be developed. This procedure is based on the enhancement of the Nielsen model (Lewis and Nielsen 1970, Riccardi et al., 2016), which takes into account the stiffening effect of aggregate particles at low frequencies and high temperatures, coupled with the use of the Voigt model.

1.4 Innovative Aspects

The innovative aspects introduced in the present thesis are summarized hereafter:

1. To develop a new procedure for estimating the rheological properties of RAP binder and of the bituminous blends, composed by fresh and RAP binder, from tests on asphalt mortars and

Chapter 1

4

implement a new procedure to estimate the Performance Grade (PG) of these binders based on the Master curves;

2. To develop new approaches, based on the rheological properties of binders and mortars, for estimating the maximum amount or RAP that can be added into a mixture without compromising its performance;

3. To extend the 2S2P1D model to predict the mechanical properties of mortars from those of binders as well as to predict the properties of asphalt mixtures from those of mortars containing RAP materials;

4. To modify a phenomenological model in order to estimate rheological properties of asphalt mixtures containing RAP materials starting from those of asphalt mortars.

1.5 Practical application of the procedure

The procedure based on the Nielsen and Voigt models can be used to back-calculate the master curves of the bituminous blends composed with fresh and RAP binder and of the RAP binder itself. From the obtained master curves the high, intermediate and low critical temperatures can be determined and then, using the blending chart, the maximum percentage of RAP that can be added to a mixture without compromising its performance can be calculated.

On the other hand, the multiscale procedure can be used to predict the complex modulus and the phase angle of mixture containing RAP, starting from tests on the corresponding asphalt mortars. These predicted values of the mixture can be used as input in the Mechanistic-Empirical Pavement Design Guide (MEPDG) (AASHTO MEPDG-1, 2008) in order to examine the responses such as stress, strain and deflection of asphalt pavements.

The MEPDG is a pavement design and performance predicting method, developed by the National Cooperative Highway Research Program (NCHRP) in 2002, that using detailed traffic loading, material properties, and environmental data, allows to compute the pavement response and to predict the incremental damage over time.

The MEPDG was adopted as a pavement design guide by the American Association of State Highway and Transportation Officials (AASHTO) in April, 2011 and in order to use this procedure, pavement engineers need a quick, easy and accurate method for obtaining dynamic modulus value avoiding laboratory dynamic modulus tests on mixtures. In fact, these tests on mixtures requires a series of expensive sampling and testing equipment, experienced laboratory personnel, and a relative long waiting time before knowing the results. Therefore, the proposed procedure to back-calculate the dynamic modulus of mixtures composed with RAP materials, starting from tests on the corresponding asphalt mortars, is an efficiently alternative method to time consuming and costly tests on asphalt mixture.

Chapter 2

5

Chapter 2

2. Literature review

The literature review is focused on some models existing in literature and that will be modified or adapted in order to achieve the specific aims of the thesis.

2.1 Existing models

E.1 2S2P1D Model

The 2S2P1D (2 Springs, 2 Parabolic elements, 1 Dashpot) is a rheological model that represents a modification of the Huet-Sayegh model (Sayegh, 1965) proposed by the research team of ENTPE/France (Olard & Di Benedetto, 2003; Olard, 2003; Di Benedetto, Olard, Sauzéat, & Delaporte, 2004; Pouget, Sauzéat, Di Benedetto, Olard, 2010; Tiouajni, Di Benedetto, Sauzéat, Pouget, 2011). It includes a linear dashpot in series with two parabolic elements and with the spring of stiffness G- G0 assembled in parallel with a second spring (G0) (Figure 2).

Fig.2 2S2P1D model

At reference temperature, the 2S2P1D model expression for complex modulus is given by:

[1]

Where i is the complex number defined by i2=-1; ω is the angular frequency such that ω= 2fr and fr

is the reduced frequency; k and h are exponents such as 0<k<h<1;  is a constant; G0 is the shear

modulus when ω→0; G∞ is the glassy shear modulus when ω→ ∞; η is the Newtonian viscosity such 1 0 0 *

)

(

)

(

)

(

1

)

(

_  _{ }























i

i

i

G

G

G

i

G

_{k h} Parabolic elements

Chapter 2

6

that η= (G∞-G0)βτ; β is a constant; τ is the characteristic time (function of temperature), based on the

Time-Temperature Superposition Principle (TTSP):

𝜏(𝑇) = 𝑎𝑇 ∙ 𝜏0 [2]

where aT is the shift factor at temperature T; τ0=τ(T0) determined at the reference temperature T0.

The shift factor at a specific temperature T, aT(T), can be obtained using the Williams-Landel-Ferry

(WFL) Equation for bituminous materials:

log 𝑎

_𝑡

(𝑇) = −

𝑐1(𝑇−𝑇0)

𝑐2+(𝑇−𝑇0) [3]

where C1 and C2 are empirical constants.

In more details, as shown in Figure 3, parameter h controls the slope at low values of G’, while k governs the slope at high values of G’ in the Cole- Cole diagram.  is associated to the slope at the low temperatures/high frequencies in the master curve of the complex modulus and the height of the maximum point in the Cole- Cole diagram, while β is linked to the slope at high temperatures/low frequencies.

Fig. 3 Visualization of the parameters of the 2S2P1D model.

Therefore, seven constants (G0, G∞, , k, h, β and τ) are needed to entirely determine the linear

viscoelastic behavior of a specific material at a given temperature. For asphalt binders the experimental static modulus is close to zero and can be assumed as negligible; hence, the number of constants can be reduced to six.

The great advantage of this model is that it is valid for any bituminous material (binder, mortar, mixture) and the constants , k, h are the same for binder and the corresponding mortars and mixtures.

= G’’

Chapter 2

7

In literature the relationship between binder and mixture can be found in Olard and Di Benedetto, 2003; Di Benedetto et al., 2004 through the characteristic time parameter:

𝜏𝑚𝑖𝑥= 10𝛼𝜏𝑏𝑖𝑛𝑑𝑒𝑟(T) [4]

where α is found by fitting the experimental results and it probably depends on mix design and ageing during mixing.

Writing Equation 1 for the binder and mixture, the following expressions are obtained:

𝐺

∗ 𝑏𝑖𝑛𝑑𝑒𝑟

(𝑖𝜔𝜏

𝑏𝑖𝑛𝑑𝑒𝑟

) = 𝐺

0 𝑏𝑖𝑛𝑑𝑒𝑟

+

_{1+𝛿(𝑖𝜔𝜏} 𝐺∞𝑏𝑖𝑛𝑑𝑒𝑟−𝐺0𝑏𝑖𝑛𝑑𝑒𝑟 𝑏𝑖𝑛𝑑𝑒𝑟)−𝑘+(𝑖𝜔𝜏𝑏𝑖𝑛𝑑𝑒𝑟)−ℎ+(𝑖𝜔𝛽𝜏𝑏𝑖𝑛𝑑𝑒𝑟)−1

[5]

𝐺

∗ 𝑚𝑖𝑥

(𝑖𝜔𝜏

𝑚𝑖𝑥

) = 𝐺

0 𝑚𝑖𝑥

+

_{1+𝛿(𝑖𝜔𝜏} 𝐺∞𝑚𝑖𝑥−𝐺0𝑚𝑖𝑥 𝑚𝑖𝑥)−𝑘+(𝑖𝜔𝜏𝑚𝑖𝑥)−ℎ+(𝑖𝜔𝛽𝜏𝑚𝑖𝑥)−1

[6]

Therefore, combining Equation 1, 5 and 6 and considering that , k, h, β are the same for the binder and the corresponding mixture, the following relationship between the mixture complex modulus and the binder one can be obtained:

𝐺_𝑚𝑖𝑥∗ _{(𝜔, 𝑇) = 𝐺}

0𝑚𝑖𝑥+ (𝐺𝑏𝑖𝑛𝑑𝑒𝑟∗ (10𝛼𝜔, 𝑇) − 𝐺𝑜𝑏𝑖𝑛𝑑𝑒𝑟)_𝐺 𝐺∞𝑚𝑖𝑥−𝐺0𝑚𝑖𝑥

∞𝑏𝑖𝑛𝑑𝑒𝑟−𝐺0𝑏𝑖𝑛𝑑𝑒𝑟 [7] This equation corresponds to a negative translation along the real axis, a homothetic expansion from the origin plus a positive translation of the binder curve in the Cole-Cole plane as depicted in Figure 4.

Fig.4 Prediction of the mixture modulus from binder one in the Cole-Cole plane

Equation 7 is known as Shift-Homothety-Shift in time-Shift (SHStS) transformation, developed by Olard & Di Benedetto, 2003; Di Benedetto et al., 2004.

Chapter 2

8

E.2 Empirical Models to determine the complex modulus of mixtures from

the complex modulus of binders

In literature a lot of empirical models were developed to predict the complex modulus of asphalt mixtures from the complex modulus of the binder; two of the most widely utilized and accepted models are those developed by Christensen et al., 2003 (Hirsch model) and by Bari & Witczak, 2006 under the NCHRP 1-40D project (Witczak 1-40D model). It is important to note that the performance of this type of model varies with the type of asphalt mixtures and other volumetric properties.

E.2.1 Hirsch model

The Hirsch model is a semi-empirical method to predict the asphalt mixture modulus. The effective response is obtained by assembling the different elements composing the mixture in series and in parallel as shown in Figure 5. In literature, different versions of the Hirsch model can be found; in the present work, we focus on the version of Christensen et al. 2003, that allows to predict the modulus of Hot Mix Asphalt (HMA), |E*|, from the shear modulus of the binder, |G*|, and from the volumetric properties of the mix as shown in Equation 8:

|𝐸∗_{| = 𝑃} 𝑐[𝐸𝑎𝑔𝑔𝑉𝑎𝑔𝑔+ 3|𝐺∗|𝑏(𝑉𝑏)] + (1 − 𝑃𝑐) [𝑉_𝐸𝑎𝑔𝑔 𝑎𝑔𝑔+ (1−𝑉𝑎𝑔𝑔)2 3 |𝐺∗_{|𝑏𝑉𝑏}] −1 [8] 𝑃_𝐶 = (20+ 𝑉𝐹𝐴∙3|𝐺∗|𝑏 𝑉𝑀𝐴 ) 0.58 650+(𝑉𝐹𝐴∙3|𝐺∗|𝑏_𝑉𝑀𝐴 )0.58 [9]

where |E*| is the dynamic modulus of the asphalt mixture (psi), Eagg, Vagg are the modulus and volume

fraction of the aggregate, |G*|b, Vb are the dynamic shear modulus (psi) and the volume fraction of the

binder, Pc is the contact volume and it is an empirical factor that determines the amount of parallel or

series elements in the mixtures, VMA is the void content in mineral aggregates (%), VFA is the volume of voids in aggregates filled with binder (%), Va is the air void content (%).

Fig.5 Semi-Empirical Model proposed by Christensen et al., 2003.

Aggregate Vo ids A sph alt bind er A ggr eg ate Voids Asphalt binder

Chapter 2

9

E.2.2 Witczak 1-40D model

The Witczak 1-40D model, expressed by Equation 10 is able to predict the asphalt mixture stiffness over a range of temperatures, loading rates, and aging conditions using the dynamic shear modulus (|Gb

*

|) and phase angle (δb) of binder as input parameters.

log |𝐸∗_{| = −0.349 + 0.754 |𝐺} 𝑏∗|(6.65 − 0.032𝜌200+ 0.0027 (𝜌200)2+ 0.011𝜌4− 0.0001(𝜌4)2+ +0.006 𝜌₃₈− 0.00014(𝜌₃₈)2_{− 0.08𝑉} 𝑎− 1.06 (_𝑉𝑉𝑏𝑒𝑓𝑓 𝑏𝑒𝑓𝑓+𝑉𝑎) + +2.56+0.03 𝑉𝑎+0.71( 𝑉𝑏𝑒𝑓𝑓 𝑉𝑏𝑒𝑓𝑓+𝑉𝑎)+0.012𝜌38−0.0001(𝜌38)2−0.01𝜌34 1+𝑒(−0.7814−05785 log(|𝐺𝑏∗|)+0.8834 log(𝛿𝑏)) [10] Where |E *| is the dynamic modulus of mixture (psi), |Gb

*

| is the dynamic shear modulus of binder (psi), ρ200 is the percentage passing #200 sieve (75 μm), ρ4 is the cumulative percentage retained on #4

sieve (4.75 mm), ρ38 is the cumulative percentage retained on 3/8 in sieve (9.5 mm), ρ34 is the

cumulative percentage retained on 3/4 in sieve (19 mm), Va are the air voids (% by volume), Vbeff is

the effective binder content, δb is the binder phase angle.

E.3 Nielsen model

The Nielsen model is a rheological model that takes into account the variation of the rheological behavior of a matrix due to aggregation of particles, degree of particle dispersion and particle size; it is able to predict the stiffening effect of filler considering the physical-chemical reinforcing effects. The original formulation of the Nielsen model (Lewis and Nielsen, 1970a; 1970b), adapted to the specific case of bituminous mortars, is expressed by equation [11]:

𝐺𝑚∗ 𝐺_𝑏∗ = 1+𝐴∙𝐵∙𝑉𝑝 1−𝐵∙Ψ∙V𝑝 [11] Where: Gm *

is the complex modulus of the mortar (composed by fine aggregate particles and binder);

Gb* is the complex modulus of the binder;

Vp is the volume fraction of fine aggregate particles calculated as the ratio of the particle volume over

the composite (mortar) volume in percentage.

A, B and 𝜓 are dimensionless model parameters that are explained in detail in the following.

A is a constant that is equal to:

A = KE-1 [12]

Where KE is the generalized Einstein coefficient that is an indicator of the physical chemical

Chapter 2

10

addition, and by this way, it includes in the equation an interaction factor between particles and suspension (Shashidhar and Romero, 1998).

For spherical particles in dilute suspension, having a perfect interface (no slippage) between particles and matrix, KE is 2.5 as derived by Einstein (Einstein, 1906); when KE is 1.0, there is no bond between

particles and binder. As KE increases, there is more of a stiffening effect with the addition of particles.

The agglomeration of particles increases the Einstein coefficient; for large agglomerates with spherical particles in cubic packing, the Einstein coefficient approaches 4.77; for particles in random packing it would be 6.76.

Particles that are elongated ellipsoids or are rod like in shape also increase the Einstein coefficient. The expected value of KE as a function of the axial ratio of the ellipsoids or rods, for the case of

randomly oriented particles, such as would occur at very low rates of shear, can be found in literature (Nielsen, Landel, 1994). High rates of shear orient the rods and decrease the effective value of the Einstein coefficient (Shashidhar and Romero, 1998).

In the case of shear modulus with spherical fillers, the value of A for any Poisson’s ratio of the matrix,

νb, is:

𝐴 = 7−5𝜈𝑏

8−10𝜈𝑏 [13]

B accounts for the relative moduli of particles and binder phases and it is equal to:

𝐵 = 𝐺𝑝⁄𝐺𝑏∗−1

𝐺𝑝⁄𝐺𝑏∗+𝐴 [14]

where Gpis the modulus of particles.

The coefficient 𝜓 in Equation [11] is given by Equation [15]:

𝜓 = 1 +1−𝜑_𝜑2 𝑉𝑝 [15]

where φ is the maximum volumetric packing fraction. It is the maximum amount of particles that can be added to the matrix without the appearance of air voids; it denotes the volume-filling contribution to stiffening and different methods can be employed to determine its value. Some of them measure the maximum volumetric packing fraction in air. Since it is related to the voids in the compacted filler, it can be estimated by using the Rigden Voids apparatus. 𝜑 is calculated as the ratio between the true and the apparent volume of the particles. Theoretically, the maximum value of 𝜑 is 0.74 for spheres in hexagonal close packing, or 0.524 in cubic packing (Nielsen, Landel, 1994). Experimentally, particles pack randomly unless great care is taken to achieve such regular orientation. For random close packing, the theoretical value of 𝜑 is equal to 0.632 whilst the experimental one is equal to 0.63. Therefore, the maximum volumetric packing fraction varies with particle shape and state of

Chapter 2

11

agglomeration. Typical values for different packing of spheres and aligned rods can be found in literature (Nielsen, Landel, 1994).

Several researchers have noted however that different fillers stiffen any binder to a different extent (Dukatz and Anderson, 1980; Anderson et al., 1982; Shashidar and Romero, 1998) and this has been a limitation to the approach of measuring the filler properties in air. Crucially, voids in the compacted filler are measured in air and do not account for interactions between the binder and the filler in the mastic. Interactions of fillers in different binder types may lead to changes in different ways. For example 𝜑 may differ in different binder as a result of different levels of dispersion of the filler, or alternatively the filler may cause changes in the viscosity of the liquid phase as a result of restructuring, physic-chemical changes or other such effects. Therefore, it would be better to estimate 𝜑 with settling test in asphalt.

By fitting Equation 11 to experimental data, the parameters A and 𝜓 and therefore KE and 𝜑 can be

estimated.

In order to better understand the significance of KE and 𝜑 on stiffening potential, the stiffening ratio is

plotted versus the volume fraction of the particles varying 𝜑 and keeping KE constant, as shown in

Figure 6.

Fig.6 Effect of variation of 𝜑 on stiffening ratio, while keeping KE constant (Shashidhar et al.,1999).

The parameter 𝜑 acts as a vertical asymptote to the curves. In fact, at Vp= 𝜑 the particles touch one

another. Mortar with Vp greater than 𝜑 is not possible, since under these conditions there will be three

different material phases (fine particles, asphalt and air voids) and in such a case the Nielsen Model cannot be applied. On the other hand, plotting the stiffness curves varying KE and keeping 𝜑 constant,

Chapter 2

12

as shown in Figure 7, an increase of the slope of the curves as KE increases can be observed. Thus, this

parameter is an indicator of the rate of increase of stiffness with addition of filler particles.

Fig.7 Effect of variation of KE on stiffening ratio, while keeping 𝜑 constant (Shashidhar et al., 1999).

E.4 Voigt Model

The structure of the micro-mechanical Voigt model for a bi-phase composite material can be idealized as shown in Figure 8. The strain in each phase is the same and the equation of the model can be written as (Lakes, 2009): Gc * = G1 * V1+G2 * V2 [16] where Gc * , G1 * , G2 *

refer to the complex shear modulus of the composite, of phase 1 and phase 2 respectively; V1 and V2 are the volume fractions of phase 1 and phase 2 respectively, with the condition

that V1+V2=1.

a)

Fig. 8 Idealized scheme of the Voigt composite: a) laminar; b) fibrous

Taking the ratio of real (G’) and imaginary (G’’) parts of the complex modulus, the loss tangent of the composite is given by (Lakes, 2009):

tanδ_c=V1tanδ1+V2 G′2 G′1tanδ2

V1+G′2_G′1V2 [17]

Chapter 2

13

where tanδc, tanδ1 and tanδ2 are the loss tangent of the composite, of phase 1 and of phase 2,

respectively; and G’1 and G’2 are the real parts of the complex modulus of phase 1 and 2, respectively.

E.5 Arrhenius Model

Arrhenius, 1887 proposed a mixing rule, expressed in Equation 18, to estimate the viscosity of a two components system, further validated for asphalt binders by Davison et al., 1994 as

𝜂_𝑚𝑖𝑥=𝜂𝐴𝛼∙𝜂𝐵

𝜂𝐵𝛼 [18]

where, ηmix is the viscosity of the bituminous blend, ηA and ηB are the viscosities of the two asphalt

binders and α is the concentration of the binder A. Since there is a direct relationship between viscosity and stiffness changes, Equation 18 can be re-written for G* as following:

𝐺∗ 𝑚𝑖𝑥=𝐺 ∗ 𝐴𝛼∙𝐺∗𝐵 𝐺∗_𝐵𝛼 [19]

2.2 Master Curves

In the present Section, two different models to plot the Master curves of the different asphalt material phases are reported.

2.2.1 Christensen Anderson Maresteanu (CAM) model

Master curves provide a fundamental rheological understanding of viscoelastic materials and allow estimating the mechanical properties over a wide range of temperature ad frequency that could be realized in the field, but that are not pratical to simulate in laboratories. In the present work, in order to plot the master curves of the complex modulus and of the phase angle, the model presented in NCHRP 459, 2001 was used. This is a universal model valid for binders, mortars and mixtures. It’s composed of three equations: 𝐺∗_{= 𝐺} 𝑒∗+ 𝐺𝑔 ∗_−𝐺 𝑒∗ [1+(𝑓𝑐⁄𝑓′))𝑘]𝑚𝑒 𝑘⁄ [20] where:

𝐺𝑒∗= 𝐺∗(𝑓 → 0) is the equilibrium complex modulus, Ge*=0 for binders; 𝐺𝑔∗= 𝐺∗(𝑓 → ∞) is the

glass transition complex modulus; k and me are two dimensionless shape parameters; fc is the location

parameter with dimensions of frequency, where the Gg* and me asymptotes intercept; 𝑓𝑐′ =

𝑓𝑐(𝐺𝑒 ∗ 𝐺𝑔∗)

1 𝑚⁄ 𝑒 is the frequency where the G

e* and me asymptotes intercept; f’ is the reduced frequency,

Chapter 2

14

Figure 9 illustrates the complex modulus master curve calculated as for Equation [20].

Fig. 9 Typical representation of a Master Curve.

These parameters allow to calculate the complex modulus and phase angle master curves by using the equations [20] and [21] respectively:

[21] where: 𝐼 = { 1 𝑓𝑜𝑟 𝑚𝑖𝑥𝑡𝑢𝑟𝑒𝑠 {_{1 𝑖𝑓 𝑓′ ≤ 𝑓}0 𝑖𝑓 𝑓′ > 𝑓𝑑 𝑑} 𝑓𝑜𝑟 𝑏𝑖𝑛𝑑𝑒𝑟𝑠

The well-known Williams–Landel–Ferry (WLF) formulation is used in the model to express the temperature-shift factor aT, as expressed in Equation 3.

This model can be used to plot the master curves of the complex shear modulus G* and of the complex modulus E*.

2.2.2 Sigmoidal model

Another model frequently used to plot the master curves of the mixtures is the sigmoidal model reported in Equation 22:

log 𝐸∗_{= 𝛿 +} 𝛼

1+𝑒(𝛽+𝛾∙𝑙𝑜𝑔𝑓′) [22]

Where δ is the minimum value of E*, α is equal to log Emax-log Emin, β and γ are shape parameters and

f’ is the reduced frequency.

2 / 2 ) ' / log( 1 ) 90 ( 90 d m d d m R f f I I                     

Chapter 3

15

Chapter 3

3. Estimation of the rheological properties of RAP binders avoiding the

extraction and recovery method

In this chapter, a new procedure to estimate the rheological properties of RAP binder and of bituminous blends composed with RAP binder, avoiding the extraction and recovery method is presented. Furthermore, a procedure to back-calculate the PG grade of these binders is introduced and two different methods to determine the maximum amount of RAP that can be added to a mixture without compromising its performance at low, intermediate and high temperature are proposed.

3.1 Development of a new procedure to estimate the RAP binder properties

from results on mortar tests

In the present Section, a new procedure specifically developed in order to determine the rheological properties of the RAP binder and of the bituminous blends, composed of fresh and different RAP binder percentages, avoiding the extraction and recovery methods, is described. The procedure is based on the rheological Nielsen model (see Section E.3), specifically adapted in order to take into account the stiffening effect at low frequency and high temperature, and on the Voigt model (see Section E.4).

The proposed methodology is summarized in Figure 10 (Riccardi et al., 2016). First, mortars consisting of fresh binder mixed together with different volume fractions of a selected fine fraction of RAP (SRAP), that are the fine fraction of the RAP, passing sieve with an opening size of 0.15 mm, or with SRAP aggregate (called Burned Selected Reclaimed Asphalt Pavement, BSRAP) obtained by ignition, (particles size smaller than 150 μm) are produced (Ma et al., 2009, Riccardi et al., 2015 and 2016). Then, Dynamic Shear Rheometer (DSR) (AASHTO T315, 2012; EN 14770, 2012) tests are performed on SRAP and BSRAP mortars. Finally, in order to back-calculate the effective complex modulus and the phase angle of the blends of virgin and SRAP binders, a new approach based on the Nielsen model (Landel and Nielsen, 1993) for composite materials is used.

The original expression of the Nielsen model is specifically adapted to take into account the effect of low frequencies and high temperatures on the stiffening contribution of fine particles in the mortar. Once the rheological properties of the bituminous blends are calculated, the simple Voigt model can be used to determine the complex modulus and the phase angle of the RAP binder. The use of both

Chapter 3

16

SRAP mortar and BSRAP mortar allows to clearly identifying the actual stiffening effect of the aggregates contained in the RAP material facilitating the identification of the model parameters. The application of this approach using RAP with different binders is explained hereafter.

1

BSRAP: Burned Selected Reclaimed Asphalt Pavement

2_{SRAP: Selected Reclaimed Asphalt Pavement}

Fig. 10 Research approach flow chart (Riccardi et al., 2016). 3.1.1 Development of the Enhanced Nielsen model (N.1)

The original formulation of the Nielsen model (Equation 11) does not take into account the effect of frequency and temperature on the stiffening ratio |Gm

*

/Gb

*

|, between the complex modulus of the asphalt mortar and that of the asphalt binder. In fact, this model was used in the past at a single frequency and at a single temperature.

Performing temperature and frequency sweep tests on mortars and binders, the stiffening ratio |Gm

*

/Gb

*

|, was calculated and then plotted as function of frequency f, at a constant temperature T=20°C. In addition, |Gm

*

/Gb

*

|, was also computed with the Nielsen model, as it was used until now, and compared with the ratio obtained using the experimental data (Figure 11).

Chapter 3

17

Fig. 11 Stiffening ratio versus frequency for different SRAP particle volume fractions (Vp) in % (e.g.

Vp20) and at T=20°C (m= measured value, N= Nielsen model calculated value).

A stiffening increase for the experimental values can be observed especially at low frequency which corresponds to intermediate and high temperatures. In fact, at lower frequencies, the binder is softer and the relative stiffening effect of particles is more significant, if not dominant. In addition, such a stiffening effect depends also on the volumetric fraction of aggregate particles. Therefore, it is not surprising that a higher stiffening effect is observed for higher Vp= 60%, for which the aggregate

skeleton may lead to significant interlocking phenomena.

In order to take into account this stiffening effect at lower frequencies, the Nielsen model needs to be adapted. Leandri et al., 2015, proposed to modify the Nielsen model adding a correction factor, which is a logarithmic function of the inverse of the testing frequency, but using this method the effects of temperature and frequency are taken into account separately and moreover the regression coefficients can vary depending on the fitting algorithm chosen and on the starting values of the coefficients. Therefore, in this work the Nielsen model in its original formulation was used, but the parameters A and 𝜑were adapted.

In particular, the time-temperature superposition principle was used in order to consider the effect of temperature and frequency at the same time. The parameters A and 𝜑 were determined by non-linear curve fitting the shifted data at different volume content of fine aggregate particles to the Nielsen model, let the parameter A to vary with the reduced frequency, and imposing a single value for the parameter 𝜑. 0 5 10 15 20 25 0 5 10 15 20 |G* m |/| G* b | f [Hz] Vp20 m Vp30 m Vp40 m Vp50 m Vp60 m Vp20 N Vp30 N Vp40 N Vp50 N Vp60 N @ T=20°C

Chapter 3

18

The parameter A, that depends on the Poisson’s ratio, as shown in Equation 13, was found to decrease when the reduced frequency increases, in accordance to the trend of the complex Poisson’s ratio versus the reduced frequency observed by Di Benedetto et al., 2010.

Moreover, the maximum volumetric packing fraction 𝜑, was also calculated as the ratio between the maximum volumes of aggregate particles obtained using the Rigden Voids apparatus and the apparent volume of aggregate particles determined in accordance to EN 1097-7(2008). The average value of the particle maximum density ρcp was determined on three specimens of the compacted particles.

Therefore, knowing the apparent density of the particles, the maximum volumetric packing fraction, 𝜑, was estimated and it results in very close agreement with literature values (Nielsen, Landel, 1994). All the values of A and 𝜑 for the different materials used in the present work are reported in Section 3.1.3.

3.1.2 Description of materials and tests

One artificially aged and three real RAP sources (one from Italy, one from England and one from Germany) were used in the present study. The first one was used to develop and to validate the proposed procedure to estimate the RAP binder properties from those of asphalt mortars using the enhanced version of the Nielsen model in combination with the Voigt model; the other RAP sources were used to confirm the applicability of the procedure to real RAP sources.

The tests carried out included the characterization of the materials to be tested:

 Determination of the binder content of the fine fraction (passing #100 sieve with an opening size of 0.150 mm) of the RAP using the Soxhlet extractor method (EN 12697-1);

 Determination of the density of the fine fraction of the RAP in accordance to EN 1097-7;

 Determination of the Rigden Voids in accordance to EN 1097-4.

Once all the needed characteristics were known, two different types of mortars were produced:



SRAP mortar composed by mixing different percentage of Selected fraction of RAP (SRAP) passing sieve with an opening size of 0.149 mm (Vp=35; 50%), shown in Figure 12 a, with

fresh binder.

 Burned SRAP (BSRAP) mortars consisting of different percentages (Vp=20; 35; 50%) of fine

fraction of the RAP aggregate extracted from the SRAP, shown in Figure 12 b, with fresh binder.

Chapter 3

19

Fig. 12 a) Selected Reclaimed Asphalt Pavement (SRAP); b) Burned Selected Reclaimed Asphalt

Pavement (BSRAP)

This sieve (#100) size limit was selected to assure a consistent reliability of the test procedure. It is good practice to prepare DSR specimens with a gap (thickness) at least ten times larger than the actual maximum aggregate particle size (Liao et al., 2013); hence, larger aggregates were not considered, since this may potentially result in misleading measurements.

In Figure 13 all terms concerning the materials used in the present study are graphically represented. As shown, Burned Selected Reclaimed Asphalt Pavement (BSRAP) represents the aggregate extracted by ignition from SRAP (Ma et., al 2010). Combining BSRAP and fresh binder, BSRAP mortars are obtained, while mixing SRAP, composed of BSRAP and RAP binder, with fresh binder, SRAP mortars can be produced.

Fig. 13 Scheme of the terms used in the present study.

In more details SRAP and BSRAP mortars were produced in order to have the same aggregate skeleton considering the RAP binder that coats the aggregates, as part of the total binder (Vb) as

represented in Figure 14.

Fig. 14 Volume distribution of mortars

Chapter 3

20

All the mortars and also the fresh binder were tested with the DSR in the classical configuration of parallel plate (AASHTO T 315, EN 14770) performing frequency and temperature sweep in order to measure the rheological properties and to plot the master curve of the complex modulus and of the phase angle. For each test and each material, three replicates were tested.

3.1.2.1 Artificial RAP source

An artificially aged binder was manufactured subjecting a 50/70 Pen grade binder to Rolling Thin Film Oven Test (RTFOT) (EN 12607-1) and 2 times on Pressure Aging Vessel (PAV) (EN 14769). Then this binder, called in the following “artificially RAP binder” was used to produce an artificial SRAP composed by the fine fraction (passing sieve with an opening size of 0.149 mm) of a sand mixed with the artificially aged binder.

The density of the fine fraction of the sand was determined in accordance to EN 1097-7 and the average between three determinations results 2.735 g/cm3.

The composition of the artificial SRAP is reported in Table 1. The binder percentage by weight results 15.48% with respect to the aggregate.

Table 1. Composition of the artificially SRAP Fine fraction of sand Binder 50/70 aged Weight (g) 129.15 20.00 γ(g/cm3 ) 2.735 1.025 Volume (cm3) 47.23 19.51 %Volume 70.76 29.24

The fresh binder used was a 70/100 Pen grade.

Moreover two bituminous blends composed by the 70/100 Pen Grade and different percentages of the artificial RAP binder, that corresponds to 35 and 50 % of SRAP, were produced.

Frequency and temperature sweep tests using the DSR in the classical configuration of parallel plate with 8 mm diameter and 2 mm gap were performed on the following binders in order to determine the complex modulus and the phase angle master curves reported in the following:

 70/100 Pen Grade

 50/70 RTFO+2PAV aged

 Bituminous blends composed by 77.7% of 70/100 and 22.3% of 50/70 RTFO+2PAV; and by 58.7% of 70/100 and 41.3% of 50/70 RTFO+2PAV. These percentages of fresh 70/100 and aged 50/70 binders were used in order to recreate the same amount of fresh and RAP binder in

Chapter 3

21

the SRAP mortars corresponding to 35 and 50% of SRAP, considering the SRAP contained 15.48% of aged binder.

Frequency and temperature sweep tests were also performed on artificial BSRAP and SRAP mortars. The strain amplitude used for the binders was 0.05% while for the mortars it was 0.005%. These values were chosen, after applying a strain amplitude, in order to keep the material response in the Linear Viscoelastic (LVE) domain. The master curves of the complex modulus and of the phase angle, determined as explained in Section 2.2, of the binders are reported in Figure 15 a and b respectively, and the master curves of the BSRAP and SRAP mortars are reported in Figure 16 and 17. The parameters of the master curves are reported in Annex 1. In Figure 15, the artificial RAP binder content of 22% and 41% by weight of total binder (virgin and RAP binder) corresponds to volume fraction of aggregate particles in mortars of 35% and 50% respectively.

Chapter 3

22

Fig. 15 Master curves a) of the Complex modulus b) of the phase angle of the fresh binder, of the

Chapter 3

23

Fig. 16 Master curves a) of the Complex modulus b) of the phase angle of the BSRAP mortars.

0 10 20 30 40 50 60 70 80 90 -8 -6 -4 -2 0 2 4 6 8 δ[ °] Log f' [Hz]

Shifted Data BSRAP Vp20 Fit BSRAP Vp20 Shifted Data BSRAP Vp35 Fit BSRAP Vp35 Shifted Data BSRAP Vp50 Fit BSRAP Vp50 b)

Chapter 3

24

Fig. 17 Master curves a) of the Complex modulus b) of the phase angle of the SRAP mortars.

3.1.2.2 Italian RAP source

An Italian RAP source with two-component binders, obtained by mixing in different proportions a hard and a soft binder were used. In order to prepare the mortars to be tested with DSR, the RAP material was sieved with the #100 (0.149 mm) sieve. The asphalt content of this RAP fraction was

0 10 20 30 40 50 60 70 80 90 -8 -6 -4 -2 0 2 4 6 8 δ[ °] Log f' [Hz]

Shifted Data SRAP Vp35 Fit SRAP Vp35 Shifted Data SRAP Vp50 Fit SRAP Vp50 b)

Chapter 3

25

determined using the Soxhlet extractor and was found to be equal to 9.89% by weight of aggregate particles.

The virgin asphalt binder used to prepare the asphalt mortar specimens was obtained by mixing different percentages of a Hard (H) and a Soft (S) binder. The following three different bituminous blends were produced:

 90%Hard+10%Soft, identified as 90H+10S;

 80%Hard+20%Soft, identified as 80H+20S;

 70%Hard+30%Soft, identified as 70H+30S.

The original H binder (100H) and the binder blends were characterized with traditional tests such as Penetration grade (EN 1426) and softening point (EN1427) according to the conventional European grading system (Table 2). The softer binder S could not be characterized due to its low consistence (Kinetic viscosity of 8000mm2/s at 60°C). The Performance Grade (PG) (AASHTO M320 2010) of all blends was also determined (Table 2).

Table 2. Asphalt binders

Using these three binders, SRAP and BSRAP mortars were produced. In total 18 mixes were prepared considering three different volume percentages, Vp, of SRAP and BSRAP (20%, 40%, 60%), and the

three different percentage combinations of the H and S binders (90H+10S, 80H+20S, 70H+30S). The compositions of the different BSRAP and SRAP mortars are reported in Table 3 and 4 respectively. The percentage of BSRAP, SRAP, fresh and RAP binder in these tables are the weight percentage respect to the total mortars.

Table 3. BSRAP mortars compositions

Vp BSRAP Fresh binder

20 39.8% 60.2%

40 63.8% 36.21%

60 79.8% 20.2%

Table 4. SRAP mortars compositions

Vp SRAP Fresh binder RAP binder in the

mortar Total binder 20 43.4% 56.6% 3.6% 60.2% 40 69.5% 30.5% 5.7% 36.2% 60 87.0% 13% 7.2% 20,2% Binder 100H 90H+10S 80H+20S 70H+30S Pen 25°C (dmm) 41 44 58 75

Softening point R&B (°C) 52 51 47 44

Fraas Braking point (°C) -6 -6 -9 -10

Viscosity at 135°C (mPa s) 370 330 295 230

Chapter 3

26

The specimens were obtained by mixing the preheated aggregate particles and the binder for two hours at 140°C; this was done to allow a complete diffusion process of the fresh binder in the RAP one, as recently demonstrated in a different study (Rad et al., 2014). In order to have the same binder aging, the BSRAP mortars were obtained following the same preparation method. The mortars were identified by a 7 digit code (00-00-00-B/S) which consists of four parts: two numbers each for the two H and S binder percentages, two numbers for the volume fraction of reclaimed material and a letter, B or S, indicating BSRAP or SRAP, respectively.

Both binders and mortars were tested using the DSR in the classical parallel plate configuration with diameter of 8 mm and gap of 2 mm, at test temperatures of 0°, 10°, 20°, 30°, 40°C. Frequency sweep tests were performed at constant strain amplitude of 0.05% for the binders and 0.005% for the mortars, in the frequency range of 0.2 to 20 Hz. The imposed strain was chosen in order to keep the material response in the (LVE) and was determined through amplitude sweep tests. The master curves are reported in Figure 18 for the binders.

Chapter 3

27

Fig. 18 Master curves of a) the complex modulus and b) of the Phase angle of the different bituminous

blends.

Figures 19 and 20 show complex modulus and phase angle master curves of the different asphalt mortars containing the same percentage of SRAP and BSRAP (Vp=20%), respectively, at a reference

temperature of 20°C. In these figures the fit of the master curves obtained with the procedure described in the NCHRP 459 (2001) are reported, while all the parameters of the master curves are summarized in Annex 1. In the case of SRAP mortar, a larger complex modulus and a smaller phase angle in comparison to the BSRAP mortar can be observed as shown in Figure 19c and 20c. This is due to the stiffening effect of the RAP binder present in the SRAP material. Moreover, the complex modulus of SRAP mortars significantly decreases as the percentage of the S binder increases due to the softening - rejuvenating effect of this softer binder. An opposite trend is observed for the phase angle. 0 10 20 30 40 50 60 70 80 90 -5 -4 -3 -2 -1 0 1 2 3 4 P h a se a n g le [° ]

Log Reduced frequency [Hz]

70H+30S 80H+20S 90H+10S

Chapter 3

28

Fig. 19 Complex modulus master curves of mortars at Vp=20% for different percentages of the H and

Chapter 3

29

Fig. 20 Phase angle master curves of mortars at Vp=20% for different percentages of the H and S

binders: a) BSRAP mortars; b) SRAP mortars; c) BSRAP and SRAP mortar for 70H+30S.

In Figures 21 and 22, the fit of the complex modulus and phase angle master curves of mortars prepared with binder blend 80H+20S and different percentages of BSRAP and SRAP are presented. The complex modulus increases while the phase angle decreases as the percentage of the aggregate particles increases; this is especially evident at lower frequency where the binder presents a softer response and the relative stiffening effect due to the aggregate particle is dominant.

Chapter 3

30

Fig. 21 Complex modulus master curves of mortars at different Vp, for a constant percentage of Hard

Chapter 3

31

Fig. 22 Phase angle master curves of mortars at different Vp, for a constant percentage of Hard and

Soft binder and T=20°C: a) BSRAP and b) SRAP mortars.

3.1.2.3 English RAP source

First, the aged binder contained in RAP was extracted and recovered using the fractionating column (Figure 23) with DCM (dichloromethane) as solvent (EN 12694-4:2005). Then

,

RAP material was sieved in order to collect the fine fraction passing #100 sieve (SRAP). Part of the SRAP was used to determine the percentage of RAP binder contained in SRAP, using the Soxhlet extractor, that results equal to 12.34% by weight respect to aggregates. The resulting aggregates were used as BSRAP and the density was determined in accordance to the EN 1097-7 and results 2.837 g/cm3.

Chapter 3

32

Fig. 23 Extraction of the RAP binder using the fractionating column.

The extracted and recovered binder contained in RAP and the virgin binder have the characteristics summarized in Table 5.

Table 5. Characteristics of the extracted and virgin binder

Penetration at 25°C [1/10 mm] Softening Point [°C] Fraass breaking point [°C] Viscosity at 135°C [mPa s] Performance Grade Extracted RAP binder 8.3 71.4 +9 1827 82-10 Virgin binder 50/70 68 47.6 -8 273 70-16

Three BSRAP mortars (Vp= 20, 35, 50) and two SRAP mortars (Vp= 35, 50) were produced with the

compositions reported in Table 6 and 7 respectively.

Table 6. BSRAP mortars compositions

Vp BSRAP Fresh binder

20 40.9% 59.1% 35 59.8% 40.2% 50 73.5% 26.5%

Table 7. SRAP mortars compositions

Vp SRAP Fresh binder RAP binder in the mortar Total binder

20 45.9% 54.1% 5.1% 59.1%

35 67.2% 32.8% 7.4% 40.2%